PRMix: Primary Region Mix Augmentation and Benchmark Dataset for Precise Whole-Mouse Brain Anatomical Delineation

Advancing whole-mouse brain analysis with novel data augmentation and expert-annotated datasets for anatomical delineation

Abstract

Breakthrough research in mouse brain anatomical delineation

The mouse brain shares remarkable similarities with the human brain across various regions, making it an essential model for studying brain pathologies, synaptic diversity, and regional specialization. To advance these studies, we have curated a collection of high-resolution dual-fluorescence microscopy images, termed as dual-fluorescent mouse brain microscopy (DMBM) dataset, complemented by expert annotations of 118 subregions. This dataset provides unprecedented insights into the molecular and structural complexity of the mouse brain. However, its full potential for detailed whole-brain analysis is compromised by challenges such as boundary ambiguity and sample scarcity in existing automated segmentation methods, prompting the development of the primary region mix (PRMix) augmentation method. PRMix is specifically designed to expand these datasets, enhance the realism of synthetic data and minimize overlap in adjacent regions. Our approach, together with the curated dataset, achieves superior segmentation performance across the mouse brain compared with existing methods, setting a new benchmark in brain imaging research.

Interactive Visualization

Experience our segmentation results in real-time

Drag the slider to compare the original dual-fluorescent microscopy image with our model's delineated output.

Before - PNG Mask
After - ROI File

Primary Regions Division

Illustration of the 11 defined primary regions and the corresponding affiliations of the 118 subregions

Primary Regions Division

The mouse brain is organized into 11 primary regions, each containing multiple subregions that contribute to different functional aspects. Our comprehensive annotation system identifies and delineates 118 distinct subregions, providing unprecedented granularity for whole-brain analysis. This hierarchical organization enables more precise segmentation and better understanding of regional specialization within the mouse brain architecture.

PRMix Pipeline

Revolutionary approach to brain image augmentation

PRMix Method

The PRMix pipeline consists of three innovative modules working in synergy. First, offline hard sample mining (HSM) identifies challenging samples that require additional attention. Second, primary region sampling (PRS) strategically selects regions for augmentation based on their anatomical significance. Finally, overlap-aware augmentation (OAA) generates synthetic data while minimizing boundary artifacts between adjacent regions, ensuring realistic and effective training samples.

Code & Dataset

Open-source resources for the research community

🔬

Source Code

The complete source code for PRMix and our segmentation models is available on GitHub, including instructions for training and evaluation.

Go to GitLab
🧠

DMBM Dataset

The expert-annotated Dual-Fluorescent Mouse Brain Microscopy (DMBM) dataset is publicly available for download on Zenodo.

Download from EIDF S3 Bucket

Paper & Citation

If you use our work, please cite our paper

Kunhao Yuan, et al. (2025). "PRMix: Primary Region Mix Augmentation and Benchmark Dataset for Precise Whole-Mouse Brain Anatomical Delineation". *Journal or Conference Name*, Vol. XX, No. Y, pp. ZZZ-ZZZ.

BibTeX

@article{Yuan2025PRMix, author = {Yuan, Kunhao and Woods, Hanan and Günar, Ülkü and Dominic, Digin and Wu, Ying and Zhen, Qiu and Grant, Seth}, title = {{PRMix: Primary Region Mix Augmentation and Benchmark Dataset for Precise Whole-Mouse Brain Anatomical Delineation}}, journal = {Journal or Conference Name}, year = {2025}, volume = {XX}, pages = {ZZZ--ZZZ}, }

Authors & Contact

Get in touch with the research team

Research Team

Kunhao Yuan, Hanan Woods, Ülkü Günar, Digin Dominic, Ying Wu, Qiu Zhen, and Seth Grant,

Center for Clinical Brain Science, The University of Edinburgh, UK

For questions and inquiries, please contact:

kyuan3@ed.ac.uk
Digin.Dominic@ed.ac.uk